* Increase _db_writer_max_scan_pct. This will give the database writer more time to flush its write list. This could result in LRU chain latch contention, because the server process searching for free buffers will be searching more buffers before giving up and posting a free buffer waits event.

* Decrease write batch size threshold. This will force the database writer to flush the write list more often, increasing the likelihood of there being a free buffer at the LRU end of the LRU chain. To decrease the write batch size, decrease the instance parameter _db_large_dirty_queue. A server process cannot move a dirty buffer to the write list if the database writer is busy writing the list's buffers to disk. If a server process, while looking for a free buffer, tries to move an unpopular dirty buffer to the being-written dirty list, it will wait, posting a free buffer waits event. If the write batch size has been increased to combat db file parallel write issues, it may have been increased too much. This is unusual, but can occur.

* Find and tune physical IO SQL. Without a block read from disk, there would be little cause for a free buffer waits event. Find the top physical IO SQL. Usually there are only a few large SQL statements consuming physical IO that clearly stand out. Tune or reduce their execution rates with the objective of reducing the amount of physical IO generated.